Trust Discrimination toward Socially Dominant and Subordinate Social Groups
Niu, Jianghe, Rosenthal, Seth A., North American Journal of Psychology
Are people in socially dominant groups more trusted than people in subordinate groups? Societies tend "to be structured as systems of group-based social hierarchies" (Sidanius & Pratto, 1999, p.31), with different groups maintaining higher and lower status levels. A dominant group is characterized by a disproportionately large share of power, wealth, social status, and good health care. A subordinate group has less power, wealth, social status, and poorer health care, and is often engaged in more high-risk occupations. Hierarchical group distinctions can involve "any of an essentially infinite number of potential distinctions between groups of human beings" (p. 48). Individuals generally accept, support, and even desire, the existence of such group-based social hierarchies. In the United States, gender, education, income, race, and increasingly, linguistic group (Wong, 2006), are distinctive hierarchical social groupings.
Trust is often described as a belief or confidence about another party's integrity (i.e., reliability, predictability, and dependability) and/or benevolence (i.e., caring, good will, and positive motives, and intentions; Mayer, Davis & Schoorman, 1995; Nooteboom, 2007; Ross & LaCroix, 1996). For the purposes of this study, trust is defined as confidence in, and positive beliefs and expectations about, other people's and groups' intentions, attitudes, and behavior. Past research on differences in trust among social groups has typically focused on the level of trust that members of particular social groups exhibit. For instance, research has explored issues such as demographic differences in the belief that "most people can be trusted" (Sztompka, 1999); gender differences in trust of unknown partners (Wang & Yamagishi, 1995); and the effects of race, ethnicity, and income on the trust of personal physicians (Schnittker, 2004; see also Sheppard, Zambrane, & O'Malley, 2004).
In contrast, the current study examines differences in trust for, rather than by, key social groups. We refer to this phenomenon as trust discrimination--the difference in trust level for one specific social group versus another social group with which it is naturally coupled (i.e., trust in males versus females, wealthy versus poor people, etc.). People's general preference for dominant over subordinate groups (Sidanius, Pratto & Mitchell, 1994) suggests that trust discrimination will follow a hierarchical pattern, with socially dominant groups trusted more than their paired socially subordinate groups. Accordingly, this study examines levels of trust discrimination between five paired social groups: Whites and non-Whites, males and females, fluent English speakers and non-fluent English speakers, more-educated and less-educated people, and those with higher income and those with lower income. The first group listed for each pair generally holds a higher social position in the United States than does the second group listed (Iwasaki, Bartlett & MacKay, 2005; Knowles & Peng, 2005; Sidanius, Singh, Hetts & Federico, 1999; Vorauer, 2003). Two hypotheses are tested in this study:
(One) Each of the five socially dominant groups is trusted more than their paired socially subordinate counterpart;
(Two) The magnitude of trust discrimination varies among paired groups.
Participants and Procedures
Participants were a diverse sample of 127 U.S. citizens (63 male, 64 female). Ages ranged from an 18-20 years age group to a 66-70 years age group (median age group = 26-30 years). 75.6% of participants were White, and 24.4% of participants were non-White; 14.3% had elementary and secondary education, 54.3% had at least some college education, and 30.7% had at least some graduate education; 95% participants were native English speakers, and 5% were not.
Participants were recruited in the greater Boston area. According to the 2000 census (U.S. Census bureau, 2000), Boston is racially, ethnically and linguistically diverse; 54.5% of residents are white, 25.3% are black, and 20.2% are Hispanic, Latino, Asian, and other; also, 33.4% speak languages other than English at home. In order to ensure accessibility to a wide demographic range of participants, recruitment was conducted in public places ranging from shopping mall food courts, and coffee- and tea-shops, to public libraries and bus and train stations. Potential participants were approached by researchers and given a brief verbal explanation of the purpose of the study. All questionnaires were completed at that time, with the researcher available to answer questions if necessary. All participation was voluntary.
Participants reported levels of received trust garnered by each of ten groups. Groups included White people, non-White people, people fluent in English, people not fluent in English, males, females, more-educated people, less-educated people, people with higher incomes, and people with lower incomes. Received trust was assessed using a measure comprised of four scenarios developed to correspond to the definition of trust discussed earlier (i.e., confidence in, and positive beliefs and expectations about, other persons' and groups' intentions, attitudes, and behaviors). The scenarios were designed (a) to be meaningful to members of the general public, and (b) to represent different domains of experience that might be encountered in daily life. See Appendix A for an example of a scenario. Participants rated (1) how likely it was that group members would make the local community better, (2) how likely they (i.e., the participant) would be to choose a group member to be the final member of a jury, (3) how much money they would expect a group member to contribute to a fundraising campaign for cancer research, and (4) how credible information they heard from a group member would be. All scenarios were rated on a seven-point scale, ranging from strongly unlikely (one) to strongly likely (seven) for the first two scenarios, from donate none (one) to donate a lot (seven) for the third, and from strongly not credible (one) to strongly credible (seven) for the fourth. Ten received trust scores were then computed for each participant by calculating their mean trust rating for each social group across the four scenarios.
Internal consistency for the measure was marginally acceptable. The mean alpha across the ten rated groups was .64. The highest alpha was .70 for trust ratings of both males and females; the lowest alpha was .51 for trust ratings of non-Whites.
Descriptive statistics and mean differences in received trust between paired groups are presented in Table 1. For all five pairs, there was a significant (p < .001) difference in received trust between the groups. In four of the cases, the socially dominant group received more trust than did the socially subordinate group: Received trust of Whites was greater than that of non-Whites; received trust of fluent English speakers was greater than that of non-fluent English speakers; received trust of more- educated people was greater than that of less-educated people; and the received trust of people with higher income was greater than that of people with lower income. In contrast, received trust of females was greater than that of males.
Comparison of the effect sizes of the mean differences between coupled groups indicates that there was stronger trust discrimination for two of the paired groups, education and English fluency, than for the other three. Specifically, the 95% confidence intervals of Cohen's d for the magnitude of the differences in received trust of those with more education over those with less education (d = 1.66) and of fluent English speakers over non-fluent English speakers (d = 1.35), did not overlap with the 95% confidence intervals for the differences between any of the other coupled groups. This indicates that the magnitude of trust discrimination was higher for education and English fluency than for race, gender, and income.
Trust Discrimination Distribution Asymmetries
We calculated difference scores for each respondent's trust ratings for each pair of coupled social groups to determine which group, if either, they trusted more. Table 2 indicates the percentages, for each coupled pair, of participants who reported more trust for the group that was trusted most on average, more trust for the group that was less trusted on average, and who trusted both groups equally. Results indicate that the trust discrimination distributions were asymmetrical; the proportion of those exhibiting more trust for one group than the other varied among the paired groups. Those with more education were trusted more than those with less education, and fluent English speakers were trusted more than non-fluent English speakers, by a higher percentage of participants (85.7% and 82.5% respectively), than were females over males (58.3%), those with higher incomes over those with lower incomes (58.3%), and Whites over non-Whites (49.2%); Z [greater than or equal to] 4.13, p < .001. Results indicated a mirror-image pattern for the groups that were less trusted on average. Non-fluent English speakers and those with less education were trusted more than their coupled counterpart groups by a smaller percentage of participants (3.2% and 4.8% respectively) than were those with lower incomes (22.8%), non-Whites (18.3%), and males (12.6%); Z [greater than or equal to] 2.01, p < .05. Finally, Whites and non-Whites, as well as females and males, were trusted equally by a higher proportion of the sample (32.5% and 29.1% respectively) than were the groups paired by income (18.9%), English fluency (14.3%), and education level (9.5%); Z [greater than or equal to] 1.76, p < .05.
Results of the study indicate that a diverse sample of participants exhibited trust discrimination for coupled groups associated with five key variables: race, language fluency, gender, education, and income. Specifically, participants trusted Whites more than non-Whites, fluent English speakers more than non-fluent English speakers, females more than males, those with more education more than those with less education, and those with higher income more than those with lower income.
In four of the five instances, hypothesis 1 was supported. The socially dominant groups (i.e., Whites, fluent English speakers, those with more education, and those with higher income) were trusted more than were their matched socially subordinate groups (i.e., non-Whites, non-Fluent English speakers, the less education, and those with lower incomes). However, in one case, the socially subordinate group (females) was trusted more than the socially dominant group (males). Interestingly, females and males were trusted equally by nearly 30% of the sample--more than any pair other than Whites and non-Whites. Additional research that examines the combined effects of both trust recipient's gender, as in the current study, and trust giver's gender (see Foubert & Sholley, 1996; Maddux & Brewer, 2005) may be instructive in understanding why trust discrimination appears to operate differently for gender than for other types of traits.
Magnitude of Trust Discrimination
Hypothesis 2 was also supported; the magnitude of trust discrimination varied among paired groups. Trust discrimination was stronger for education level and English fluency than for the other paired groups (i.e., race, gender, and income). Interestingly, two of the three paired variables with the lowest levels of trust discrimination, race and gender, are inborn traits that are generally regarded as fixed. In contrast, the two traits for which there was the greatest trust discrimination, education and English fluency, are generally regarded as environmental. Although this does not explain the trust discrimination results for income, further research may help determine whether trust is generally lower for members of socially subordinate groups who are believed to have some personal responsibility for their social status than for members of socially subordinate groups where membership is innate.
With respect to English fluency levels it is particularly important to note that increasing numbers of people in the United States speak a language other than English. According to the 2000 Census (U.S. Census Bureau, 2000), there are forty-seven million people in the United States who speak other languages at home. In some states, such as Hawaii and New York, about one-third of the residents speak other languages more fluently than they speak English. In California, about 40% of the people are not fluent English speakers. Such situations can lead to separation between large segments of the population through assortive socializing based on English fluency levels (Wong, 2006). Distrust of social groups in the U.S. based on English fluency may be the source of discrimination, and may cause painful emotional responses and decreased self-confidence among the large number of non-fluent English speakers who live and work in the U.S. (McCormack, 1998).
One major limitation of our study is that the sample was biased against some groups. For example, 95% of the participants were native English speakers. If this percentage had been closer to the norm for the Boston area perhaps the effect for non-fluent English speakers would have been smaller. Almost 31% of the participants had taken some graduate level courses. This figure suggests that the sample had somewhat more education than a random sample of adults, and might account for some of the lower trust shown toward those with less education. The fact that whites were trusted more than non-whites may reflect the fact that 76% of our sample was white.
It is also not clear in retrospect that responses to the four seven-point Likert scales are pure measures of trust. Consider the question "How much money would you expect a low income person to contribute ..." If a respondent chooses "donate none" does that mean that he/she doesn't trust low income persons? Or, does it merely mean that she/he thinks the low income person has little to donate? Question four asks about the credibility of information received. Aren't highly educated persons usually a better source of information than persons with little education? Do low scores directed toward the latter reflect lack of trust, a belief that education works, or some combination of the two?
For future research on this topic we recommend the use of different samples for each of our five variables, samples that are randomly selected and about equal in numbers, with analyses for each group separately as well as together. We also suggest a refinement of the present measure of trust and the use of two or more additional measures of trust to tap the idea that trust has several possible meanings.
Summary and Conclusion
The current research suggested that, in all cases but one, participants trusted the group with higher social status more than the associated group with lower social status. However, they also trusted one group with lower social status, females, more than the associated group with higher social status, males.
Magnitude of trust discrimination also varied among paired groups. Trust was lowest for members of subordinate groups who may be perceived as having personal responsibility for their social status--those with less education and non-fluent English speakers.
Fukuyama (1995) and Putnam (2000) believed that trust should not be given blindly; authentic trust should be earned. Brewer (1999) further suggested that indiscriminate trust is not an effective individual strategy. In other words, it is not necessarily a given that people should trust others equally. However, knowledge about which social groups are most and least trusted, and why that may be the case, is crucial in helping to uncover prejudices that pervade society. Understanding trust discrimination may then help us anticipate and prevent the types of behavioral discrimination toward social groups for which low levels of trust may be responsible.
(EXAMPLE OF A SCENARIO)
One more JUROR is needed for hearing a case. How likely would you be to choose that juror from each of the following [ten] groups? Please circle the number (1 = Strongly Unlikely, 2 = Unlikely, 3 = Slightly Unlikely, 4 = Neutral, 5 = Slightly Likely, 6 = Likely, 7 = Strongly Likely) that best expresses your opinion about each group.
Author Notes: Jianghe Niu, Department of Psychology, Harvard University. Seth A. Rosenthal, Center for Public Leadership, Harvard Kennedy School. Heartfelt thanks to Michael Dagg and Serge Arpin for supporting the research. Great appreciation is given to the following people for assisting the research and the paper: James Sidanius, Howard Gardner, Arnold Ho, Carrie James, Margaret Rundle, Terrence Tivnan, Chunglin Feng, Marilyn Hamilton, Andy Hamilton and Mu-Qing Huang. Copies of all scenarios are available from the authors.
Foubert, J. D., & Sholley, B. K. (1996). Effects of gender, gender role, and individualized trust on self-disclosure. Journal of Social Behavior & Personality, 11, 277-288.
Iwasaki, Y., Bartlett, J., & MacKay, K. (2005). Social exclusion and resilience as frameworks of stress and coping among selected non-dominant groups. International Journal of Mental Health Promotion, 7, 4-17.
Knowles, E. D., & Peng, K. (2005). White selves: Conceptualizing and measuring a dominant-group identity. Journal of Personality and Social Psychology, 89, 223-241.
Maddux, W. W., & Brewer, M. B. (2005). Gender differences in the relational and collective based for trust. Group Processes & Intergroup Relations, 8, 159-171.
Mayer, R. C., Davis, J. H., & Schoorman, F. D. (1995). An integrative model of organizational trust. Academy of Management Review, 20(3), 709-734.
McCormack, A. S. (1998). Revisiting discrimination on campus: 1988, 1992, and 1996. College Student Journal, 32, 378-393.
Nooteboom, B. (2007). Social capital, institutions and trust. Review of Social Economy, 65(1), 29-53.
Ross, W. & LaCroix, J. (1996) Multiple meanings of trust in negotiation theory and research: A literature review and integrative model. The International Journal of Conflict Management, 7, 314-360.
Schnittker, J. (2004). Social distance in the clinical encounter: Interactional and sociodemographic foundations for mistrust in physicians. Social Psychology Quarterly, 67, 217-235.
Sheppard, V. B., Zambrana, R. T., & O'Malley, A. S. (2004). Providing health care to low-income women: A matter of trust. Family Practice, 21, 484-491.
Sidanius, J., & Pratto, F. (1999). Social Dominance: An Intergroup Theory of Social Hierarchy and Oppression. Cambridge, United Kingdom: Cambridge University Press.
Sidanius, J., Pratto, F., & Mitchell, M. (1994). In-group identification, social dominance orientation, and differential intergroup social allocation. Journal of Social Psychology, 134(2), 151-167.
Sidanius, J., Singh, P, Hetts, J. J., & Federico, C. (1999). It's not affirmative action, it's the blacks: The continuing relevance of race in American politics. In D. O. Sears, J. Sidanius, & L. Bobo (Eds.) Racialized Politics. Chicago: The University of Chicago Press, Ltd. London.
Sztompka, P. (1999). Trust: A Sociological Theory (pp. 102-118). Cambridge, United Kingdom: Cambridge University Press.
U.S. Census Bureau. (2000). Language spoken at home and ability to speak English for the population 5 years and over by state: 200. Census 2000, Summary File 3, Table p19.
Vorauer, J. D. (2003). Dominant group members in intergroup interaction: Safety of vulnerability in numbers? Personality and Social Psychology Bulletin, 29, 498-511.
Wang, F., & Yamagishi, T. (1995). Group-based trust and gender differences in China. Asian Journal of Social Psychology, 9, 199-210.
Wong, S. (2006). Perpetual foreigners: Can an American be an American? In A. Curtis & M. Romney (Eds.). Color, Race, and English Language Teaching: Shades of Meaning (pp. 81-92). Mahwah, NJ. US: Lawrence Erlbaum.
Author info: Correspondence should be sent to: Dr. Jianghe Niu, Department of Psychology, Harvard University, Cambridge, MA 02138
Jianghe Niu & Seth A. Rosenthal
TABLE 1 Received Trust Means and Standard Deviations; Mean Differences and Effect Sizes for Paired Groups Mean SD t Cohen's d 95% Confidence Interval of d White 4.62 .86 4.01 * 0.31 0.16 - 0.45 Non- 4.36 .80 White Fluent 4.97 .83 12.77 * 1.35 1.21 - 1.51 English Non-fluent 3.80 .90 English Male 4.46 .87 -5.03 * 0.43 0.27 - 0.58 Female 4.84 .90 More- 5.31 .84 14.09 * 1.66 1.52 - 1.83 educated Less- 3.83 .94 educated Higher- 4.67 .95 4.24 * 0.45 0.28 - 0.60 income Lower- 4.26 .88 income Note: N = 127. * p < .001. TABLE 2 Percent Favoring More Trusted Group, Less Trusted Group, or Favoring Both Groups Equally More Trusted Less % More % Less % Groups Group Trusted Trusted Trusted Favored Group Group Group Equally Favored Favored White Non-White 49.2 18.3 32.5 Fluent English Non-fluent 82.5 3.2 14.3 English Female Male 58.3 12.6 29.1 More-educated Less-educated 85.7 4.8 9.5 Higher-income Lower-income 58.3 22.8 18.9 Note. N = 127…